Temptations and dynamic consistency

Temptations and Dynamic Consistency Enrica Carbone 1 Università di Bari Facoltà di Economia Via Camillo Rosalba 53 e-mail:[email protected]

ABSTRACT

The objective of this paper is to test a prediction of the quasi-hyperbolic model. The test is innovative in that it uses an experimental implementation in which there are two treatments: a forward market and a spot market. In each of these markets goods and activities are sold. The good and activities sold are investment goods or activities and temptation goods or activities. The prediction of the quasi-hyperbolic model is that in the spot (forward) market participants will buy more temptation (investment) goods and activities than in the forward market and less investment (temptation) good and activities. This prediction is not confirmed by the data, and hence is at odds with previous experiments which have shown support for the hyperbolic model. We speculate on the reasons for this.

1

The author thanks David Laibson for stimulation and discussions.

1

1. Introduction The objective of this paper is to test an implication of the quasi-hyperbolic model of discounting by implementing an experiment on temptations. This implication is that people choose more investment goods than temptation goods when they plan, but choose more temptation goods than investment goods when they consume on the spot. This effect, which has been called the immediacy effect by Read, Loewenstein and Kalyanaraman (1999), occurs when an individual behaves according to the quasi-hyperbolic discounting model, but not when an individual behaves according to the exponential discounting model. The immediacy effect has been tested by Read et al (1999) and by Read and Barbara Van Leeuwen (1998) in a particular experimental setting. These authors conclude from their experiments that the immediacy effect exists. However, the results of the experiment reported in this current paper – in a different experimental setting - do not confirm this effect. The main point of this current paper is not only to show that the immediacy effect is not observed in our experiment, but also to compare our results to those of Read et al, and try to understand why they are different. We will argue that these differences are due to important and relevant differences in the experimental design. The paper is organised as follows. In the next section we briefly outline the exponential and quasi-hyperbolic models of discounting. We then apply these two theories in a particular context and show that they may lead to different predictions. In the third section we describe our experiment which was designed to try and distinguish whether individuals behave according to the exponential model or according to the quasi-hyperbolic model. The fourth section contains an analysis of the data analysis and the final section concludes.

2. The Theory Tested

2

The mainstream model of intertemporal choice in the economics literature is the Discounted Utility Model - in which the utilities of future consumption are discounted to the present. The Discounted Utility Model is typically implemented with an exponential discount function. According to this exponential discounting model, an individual discounts the future utility of consumption using a constant discount factor δ, so that the utility at time t of a stream of consumption ct, ct+1, … , cT is given by the expression: u (ct ) + δ u (ct +1 ) + δ 2u (ct + 2 ) + ... + δ T −t +1u (cT )

(1)

Phelps and Pollak (1968), introduced the concept of quasi-hyperbolic discounting. This has recently been ‘re-discovered’ and extended by Laibson (1997). The quasi–hyperbolic discounting model is built on the idea (reinforced by empirical evidence) that consumers have a higher discount rate between the present and the following period than between any two adjacent subsequent periods. This quasi-hyperbolic discounting model implies that the utility at time t of a stream of consumption ct, ct+1, … , cT is given by the expression: u (ct ) + β [δ u (ct +1 ) + δ 2u (ct + 2 ) + ... + δ T −t +1u (cT )]

(2)

The existence of the parameter β (if it is not equal to 1) distinguishes this model from the exponential model. Note that, as viewed from period t, the individual discounts the utility of period t+s+1 consumption relative to the utility of period t+s consumption by δ, whereas, as viewed from period t+s, the individual discounts the utility of period t+s+1 consumption relative to the utility of period t+s consumption by βδ. Thus the relative discount rate varies according to the time of the comparison. This implies the possibility of time inconsistency. The hypothesis we test is an implication of the hyperbolic model combined with the use of temptation goods or activities and investment goods or activities. Let us define an investment good or activity as a good which has current costs and future benefits, while a temptation good or activity is one that has current benefits and future costs. Read el al (1999) use the terms vices and virtues: their virtues are our investment goods and activities; and their vices are our temptation goods and activities. Read et al note (p 259), “Someone who discounts the future in a hyperbolic or quasi3

hyperbolic manner will be likely to prefer an immediate vice over an immediate virtue, since the vice offers a larger reward in the present. The same individual, however, might well take the virtue if both are delayed, since in this case the initial reward offered will no longer receive disproportionate weight”. This is the hypothesis that we test in our experiment. Let us state this more formally. Suppose that in the case of a temptation good the cost is paid a period after the benefit is received, while in the case of an investment good the benefit is received the period after the cost is paid. If an individual with hyperbolic discounting has to choose in the present, then, if he or she chooses the temptation (investment) good, the present benefit (cost) will not be discounted, while the future cost (benefit) will be discounted by βδ. However, if the same individual has to decide now about what to consume in the following period, then, if he or she chooses the temptation (investment) good, then the present benefit (cost) will be discounted by βδ, while the future cost (benefit) will be discounted by βδ2. The relative discounting changes according to the time that the decision is made. To illustrate this with a concrete example, we follow one given by Read et al. Assume that β = 0.5 and δ = 1, and that the utility stream is (25, 200) for the virtue and (100, 100) for the vice, where the first entry represents the utility received when the vice or virtue is consumed and the second entry the utility in the period following. The individual should choose according to the discounted values of the two streams. If consumption follows immediately after the decision then the relevant discounted utilities are 125 ( = 25 + 0.5*200) for the virtue and 150 ( = 100 + 0.5*100) for the vice. The vice will be preferred. On the other hand, if the consumption can only take place one period after the decision, then the relevant discounted utilities are 112.5 ( = 0 + 0.5*(25+200)) and 100 ( = 0 + 0.5*(100+100)). Now the virtue will be preferred. We now apply these ideas in a different context. Consider two different market situations: the first in which there is a spot market; and the second in which there is a forward market. In both markets good and activities are sold. In the spot market participants can consume at the same time as they buy. However, in the forward market, participants have to order the goods in advance of 4

their consumption. The effect described above, called by Read et al the immediacy effect, leads to the following prediction concerning differences in behaviour between the spot market and the forward market: people will chose more investment (temptation) goods or activities in the forward (spot) market than in the spot (forward) market. Of course, as Read et al say, the possibility of observing this prediction depends on the relative desirability of the temptation and investment goods and activities.

3. The Experimental Design Read et al’s experimental setting required subjects to attend on two occasions. On the first occasion subjects were asked what they would like to consume (from a list of possibilities) when they returned on the second occasion. When they did actually return on the second occasion, they were asked again what they wanted to consume (from the same list) on that occasion. Note that the responses of the subjects on the first occasion were not implemented, so that subjects could well – and in fact did – change their minds. Indeed, that was the crucial finding of the experiment: de facto on the second occasion subjects consumed more vices than they had said (on the first occasion) that they wanted to consume on the second occasion. Read et al take this as confirmation of the immediacy effect. For reasons that we discuss later, we prefer a different experimental setting. This is described below and takes its cue from the application of the quasi-hyperbolic model to the case of spot and forward markets discussed at the end of the section above. The experiment was conducted at ESSE (Economia Sperimentale al sud d’Europa) Lab at the University of Bari in Italy. The experiment consisted of eight sessions; four sessions were run at the end of January 2005 and the other four at the end of May 2005. The experiment was advertised through a leaflet distributed by hand in the Faculty of Economics at the University of Bari or sent by e-mail to a list of people that had participated in previous experiments. This leaflet (available on request) informed people that the experiment would last 5 hours, that the participants could bring with them their textbooks, that during the experiment they could read magazines, play videogames 5

and so on, and at the end of the experiment they would receive 50 euros less what they had spent during the experiment. There were two separate treatments in the experiment: a spot market treatment and a forward market treatment. Four of the sessions used the spot market treatment and four the forward market treatment. In each treatment a range of investment goods and activities and a range of temptation goods and activities were available for sale 2 . In the spot market treatment the participants had 15 minutes to read the instructions (simultaneously read aloud by the experimenter) and subsequently 4 hours and 45 minutes to buy and consume goods and activities that they wanted. In the forward market treatment the participants had 15 minutes to read the instructions (simultaneously read aloud by the experimenter) and order the goods and activities they wanted to consume during the subsequent 4 hours and 45 minutes; crucially no other goods or activities could be bought after the first 15 minutes and the subjects were fully informed that that would be the case. During the ensuing 4 hours and 45 minutes the subjects were brought the goods and activities that they previously had ordered. In neither treatment could participants talk to each other during the experiment. Nor could they indulge in any other activity (other than doing nothing) other than those available in the experiment; nor could they consume any other good other than those available in the experiment. At the end of the experiment the participants were paid 50 euros less what they spent during the experiment; All the goods were sold at half price of the faculty bar prices. All the activities were sold at 1 centesimo (of euro) per minute. Goods and activities bought during the experiment, but not consumed, could not be taken away or refunded at the end of the experiment. At the end of the experiment participants answered a questionnaire, which allowed us to elicit some demographic information. Additionally in the questionnaire each subject had to rate each good in term of immediate pleasure and delayed benefit. We used these ratings to calculate temptation indices and investment indices for the various goods and activities (both individual – to 2

The complete list of goods and activities is in the instruction reported in Appendix 1

6

the subject – and aggregate – to the participants as a whole). We shall say more about these indices shortly. The consent form, the instructions, the debriefing questionnaire and the debriefing statement are reported in Appendix 1.

4. The Data Analysis Recall that the prediction of our model is that people with hyperbolic discounting would consume more temptation goods in the spot market than in the forward market, and consume less investment goods in the spot market than in the forward market. We now proceed to a test of this prediction. We describe the nature of our empirical investigation, after introducing some notation. In this, we refer to individual goods and activity with the subscript i and individual subjects with the subscript j. In our experiment there was a total of 24 goods and activities so i ranges from 1 to 24 and there was a total of 80 subjects so j ranges from 1 to 80 . The variables we observed during the experiment are the following: eij: the expenditure on good and activity i by individual j. From these we can derive the total expenditure on good or activity i: ei =

80

∑e j =1

activities of all individuals e =

ij

and the total overall expenditure on all goods and

24

∑e . i =1

i

Tij: the temptation felt by individual j with respect to good or activity i. This was measured by the response of the individual in the questionnaire to the question “Does this good or activity give you immediate pleasure? Is it fun, tasty or pleasurable”. The variable Tij was coded 0, 1 or 2 according as the response was “none”, “a little” or “a lot”. Iij: the “investability” felt by individual j with respect to good or activity i. This was measure by the response of the individual in the questionnaire to the question “Does this good or activity have 7

benefits that last at least a few days? Is it healthy or educational?”. The variable Iij was coded 0, 1 or 2 according as the response was “none”, “a little” or “a lot”. This measures how much investment value there was to that individual of that good or activity. Ti and Ii: average measures of temptation and investability of each good or activity defined as follows: Ti =

1 80 1 80 Tij and Ii = ∑ ∑ Iij 80 j =1 80 j =1

F: a dummy for the forward market (forward F=1 and spot F=0). In addition we gathered the following demographic data: whether the subject was a student; whether the subject was employed; the subject’s annual food expenditure; his or her age; his or her height; his or her gender; his or her weight. The first equation estimated is the following: eij = 0.002 + 0.084Tij + 0.117Iij +0.174 Fj – 0.053 Fj Tij -0.147 Fj Iij (0.085) (4.962) (6.459) (6.677) (-2.395) (-6.325)

(i = 1..24, j = 1..80)

Note that the above equation implies: Spot market (F = 0):

e = 0.002 + 0.084 T + 0.117 I

Forward market (F = 1):

e = 0.176 + 0.031 T – 0.030 I

When we substitute in the 3 possible values of the temptation index and the 3 possible values of the investment index we obtain the following table:

T 0 1 2 0 0 0

Spot market I 0 0 0 0 1 2

e 0.002 0.086 0.170 0.002 0.119 0.236

T 0 1 2 0 0 0

Forward market I 0 0 0 0 1 2

e 0.176 0.207 0.238 0.176 0.146 0.116

This table tell us that when the temptation index is positive (1 or 2) and the investment index is zero that expenditure in the forward market is higher than the expenditure in the spot market. This goes

8

against the predictions of the hyperbolic model. However, when the investment index is 1(2) and the temptation index is zero the expenditure in the spot market is lower (higher) than the expenditure in the forward market – so this evidence partly supports the quasi-hyperbolic prediction and partly refutes it. An alternative, and perhaps more direct and robust, way of analysing the data is the following. Let us weight expenditures by their temptation and investment content. We have the temptation/investment coefficients from the questionnaire and depending on how we use them we can calculate weighted expenditure on temptation and investment goods and activities. Ideally we would like to say that expenditure on a particular good or activity is x% temptation and (100-x)% investment – with the obvious extremes 100% temptation and 100% investment. The first of these occurs when the subject replies “a lot” to the temptation question and “none” to the investment question, and the second when the subject replies “none” to the temptation question and “a lot” to the investment question. Also it is clear when the subject replies “a little” to both questions that the good or activity is 50% temptation and 50% investment. However other cases are not so clear, and we have to decide a mapping from these answers on the questionnaires to the temptation and investment content of the expenditure. Consider the following mapping: Temptation coefficients 1/2 0 1 2 0 0.00 0.00 0.00 1 0.50 0.50 0.50 2 1.00 1.00 1.00

Investment coefficients 0 1 2 0.00 0.50 1.00 0.00 0.50 1.00 0.00 0.50 1.00

\ We note that, with this mapping, if the temptation index is 0, 1, 2 then the expenditure is 0%, 50%, 100% temptation, irrespective of the investment index. It the investment index is 0, 1, 2 then the expenditure is 0%, 50%, 100% investment, irrespective of the temptation index. However if both indexes are 2 this mapping implies that the expenditure is allocated 100% to temptation and 100% to investment – and hence is double counted. Clearly there is no ‘correct’ mapping, particularly if an individual regards a good or activity as providing both immediate and delayed pleasures. However, using this mapping (at an individual level using individual responses to the questionnaire 9

to weight the individual expenditures) we get the following implied weighted expenditure percentages: Temptation Investment Spot

58

42

Forward

59

41

This shows that the expenditure on temptation goods and activities in the forward market is slightly higher than in the spot market, and that the expenditure on investment goods and activities is slightly higher in the spot then in the forward market. This is the opposite of that predicted by the quasi-hyperbolic model. Of course, we may have chosen a bad mapping, but the above result appears robust. For the record, in Appendix 2, we include other mappings and their implied weighted expenditures. The implications do not seem particularly sensitive to the mapping.

5. Conclusions The prediction of the quasi-hyperbolic model we tested in this experiment is not confirmed by the data. This prediction is that people choose more investment goods than temptation goods when they plan, but choose more temptation goods than investment goods when they consume on the spot. To detect the plan that people make, and to compare that with what people consume on the spot, we used an experimental design in which there were two treatments: one containing a forward market and the other a spot market. We do not observe the expenditure reversal predicted by the hyperbolic model. What is observed, on the contrary, is that expenditure on temptation goods in the forward market is higher than in the spot market, and that expenditure on investment goods is lower in the forward market respect than in the spot market. However, we should not dismiss immediately the quasi-hyperbolic discounting model on the basis of these results. There are two features of the experiment that may have caused this contrary 10

result: one feature is the delay between the moment that people choose and the moment that people consume. One could argue that the probability that the preference reversal happens is inversely related to the temporal distance between the decision and the moment of consumption. In this experiment people would implement their plan soon after having chosen (they had a quarter of an hour to read the instruction and choose). We could, in order to put more time between the decision and the implementation of the plan, have implemented an experimental design in which the participants in the experiment had to go away and return on a second occasion. This would have created other problems – not least that the participants could have prepared for the second experimental occasion by eating and drinking in advance and coming with prepared reading. The second problem in our experimental design is that which might be called the commitment or insurance effect: if people are risk averse and have to decide in advance what they are going to consume, they may choose more goods that they really want to avoid ending up without doing anything or eating less that they would desire. So they ordered more goods and activities they wanted - simply as insurance. It is not clear how one might control for this effect. However, our design has advantages over that of Read et al. In their design, there appear to be two problems, both related to the decision at the second stage. First, and de facto, subjects were allowed to change their minds on the second occasion – so that any plan that they formulated could be changed. Second, and as a consequence of this, if the subjects knew on the first occasion that they would be allowed to change their minds on the second occasion, it is not clear in what sense their stated plans at the first stage were in fact their true plans. However, and this is still something to be explained, our experiment shows that we do not find any confirmation of the prediction of the quasi-hyperbolic model. We might have expected to find a modest effect, but, in fact, the movement of expenditure goes in the wrong direction.

11

6. References Carbone E., (2006), “Understanding Intertemporal Choices”, Applied Economics, Vol. 38. Harrison G.W., Lau M.I. and Melonie W.B., (2002), “Estimating Individual Discount Rates in Dennmark: A Field Experiment”, The American Economic Review, Vol.92, n.5. Laibson D., (1997), “Golden Eggs and Hyperbolic Discounting”, The Quarterly Journal of Economics, Vol.112, n.2. Phelps E.S. and R. A. Pollak (1968), “On Second-Best National Saving and Game-Equilibrium Growth”, Review of Economic Studies, 35, 165-80. Read D. Van Leeuwen B., (1998), “Predicting Hunger: The effects of Appetite and Delay on Choice”, Organizational Behavior and Human Decision Processes, Vol.76, n.2. Read D. Loewenstein G. and Kalyanaraman S., (1999), “Mixing Virtue and Vice: Combining the Immediacy Effect and the Diversification Heuristic”, Journal of Behavioral Decision Making, Vol.12, n.4.

12

Appendix 1: the consent form, the instructions and the debriefing questionnaire and statement Consent form: Professor Enrica Carbone of the Faculty of Economics at The University of Bari is conducting a study on how people make decisions in real-world environments. You may participate in this study only once. You will receive 50€, which you can spend on food or activities during the experiment. Whatever you don’t spend you can take home with you at the end of the experiment. Your total participation should take around 5 hours. The data collected in this study will be used for economics research. Though nothing in the experiment is personally sensitive, we have taken steps to assure your anonymity. This anonymous data will be analyzed in future research. I have read these instructions and agree to participate in the study: ____________________________________ (signature and date).

13

I. Instructions for the Experiment This experiment lasts 5 hours. During the first fifteen minutes we will explain the instructions and show you the goods and activities that can be purchased. During this time you will also be able to select the goods and activities that you want to purchase. After this, you will have 4 hours and 45 minutes to use the goods and activities that you purchased. Goods and activities purchased but not consumed during the course of the experiment cannot be taken with you at the end of the experiment or refunded. The selection and purchase of goods and activities can only be done from 12.30 to 12.45. To purchase a good or an activity, fill in the quantity on this page and the next page. After 12.45 no changes can be made. During the experiment, participants can speak only with the experimenter and her assistants. Please do not talk to other participants. The prices for goods and activities are listed below. If you order something (like a cup of espressino), it will be delivered to you whenever you ask for it during the experiment. For example, if you purchase two cups of coffee, you have purchased the right to tell the experimenter to bring you a cup of coffee twice during the experiment. Remember that goods and activities purchased but not consumed during the course of the experiment cannot be taken with you at the end of the experiment or refunded. We first list food items from the bar at the Faculty of Economics. Write the number you want (or “0”). • Espressino coffee 24 centesimi: I agree to purchase ____ cups of espressino. • Espresso coffee 23 centesimi: I agree to purchase ____ cups of espresso. • Chocolate Snack 50 centesimi: I agree to purchase ____ chocolate snacks. • Focaccia 90 centesimi: I agree to purchase ____ focaccia. • Sandwich 90 centesimi: I agree to purchase ____ sandwiches. • Oransoda 40 centesimi: I agree to purchase ____ oransodas. • Lemosoda 40 centesimi: I agree to purchase ____ lemonsodas. • Chinotto 40 centesimi: I agree to purchase ____ Chinotto. • Crostini 35 centesimi: I agree to purchase ____ crostini. • Salad (lettuce, rucola, tomato, raw ham, salt and a teaspoon of oil served with bread) €1,50: I agree to purchase ____ salads. • Fruit 40 centesimi: I agree to purchase ____ pieces of fruit. • Yoghurt low fat 60 centesimi: I agree to purchase ____ cups of yoghurt. • Crisps 35 centesimi: I agree to purchase ____ crisps. • Cucciolone 90 centesimi: I agree to purchase ____ cucciolone • Magnum Nuts 90 centesimi: I agree to purchase ____Magnum Nuts • Cornetto €1: I agree to purchase ____Cornetto • Caramelle 60 centesimi: I agree to purchase ____ caramelle. Non-food items: you will be charged by the minute; write the total number of minutes you want (or “0”). • Textbook reading 1 centesimo per minute: I agree to purchase ____ minutes of textbook reading. • Research on the catalogue of the departmental library 1 centesimo per minute: I agree to purchase ____ minutes of research on the catalogue of the departmental library. • Research on the electronic catalogue of the university library 1 centesimo per minute: I agree to purchase ____ minutes of research on the electronic catalogue of the university library. • Weights and gym equipment for developing pectoral muscles 1 centesimo per minute: I agree to purchase ____ minutes of use of the weights and gym equipment. • Magazines (Panorama, l’Espresso, Economy) 1 centesimo per minute: I agree to purchase ____ minutes of magazine reading. • Video games 1 centesimo per minute: I agree to purchase ____ minutes of video game playing. • Music on the Virgin Radio, Italian Radio web sites, and CDs 1 centesimo per minute: I agree to purchase ____ minutes of web music.

From 12.45 to 17.30 you will be able to use the goods and activities that you agreed to purchase. At 17.30 you will be given a questionnaire. After completing the questionnaire, you will be paid €50 minus what you spent on goods and activities.

14

II. Instructions for the Experiment This experiment lasts 5 hours. During the first fifteen minutes we will explain the instructions and show you the goods and activities that can be purchased. After this, you will have 4 hours and 45 minutes to purchase goods and activities. To purchase a good or an activity, simply tell the experimenter that you want something. During the experiment, participants can speak only with the experimenter and her assistants. Please do not talk to other participants. The prices for goods and activities are listed below. If you order something (like a cup of espressino), it will be delivered to you as soon as possible. We first list food items from the bar at the Faculty of Economics. • Espressino coffee 24 centesimi. • Espresso coffee 23 centesimi. • Chocolate Snack 50 centesimi. • Focaccia 90 centesimi. • Sandwich 90 centesimi. • Oransoda 40 centesimi. • Lemosoda 40 centesimi. • Chinotto 40 centesimi. • Crostini 35 centesimi. • Salad (lettuce, rucola, tomato, raw ham, salt and a teaspoon of oil served with bread) €1,50. • Fruit 40 centesimi. • Yoghurt low fat 60 centesimi. • Crisps 35 centesimi. • Cucciolone 90 centesimi. • Magnum Nuts 90 centesimi. • Cornetto €1. • Caramelle 60 centesimi. Non-food items: you will be charged by the minute. • Textbook reading 1 centesimo per minute. • Research on the catalogue of the departmental library 1 centesimo per minute. • Research on the electronic catalogue of the university library 1 centesimo per minute. • Weights and gym equipment for developing pectoral muscles 1 centesimo per minute. • Magazines (Panorama, l’Espresso, Economy) 1 centesimo per minute. • Video games on the RealArcade web site 1 centesimo per minute. • Music on the Virgin Radio, Italian Radio web sites, and CDs 1 centesimo per minute. From 12.45 to 17.30 you will be able to use the goods and activities above. At 17.30 you will be given a questionnaire. After completing the questionnaire, you will be paid €50 minus what you spent on goods and activities.

15

Debriefing Questionnaire

Please answer the following questions: How did you hear about the experiment? ________________________________________________________________________________ ________________________________________________________________ Are you a student at Bari? Yes ____ No _____. If you are not a student at Bari, what is your current employement status? ___________________. If you are a student, what is your field of study: _____. How many Euros do you spend per year on food? ________. What is your age: ____. Are you married? Yes ____ No ____. What is your height: ____ meters. Are you male or female: Male ___ Female ___. What is your weight: _____ kilograms. Were any parts of the experimental instructions confusing? Yes _____ No _____. If you answered yes, please describe which part was confusing: ________________________________________________________________________________ ________________________________________________________________

What do you think this experiment was about? Please explain below: ________________________________________________________________________________ ________________________________________________________________

16

Please rate the following goods and activities. Give each good or activity two ratings. Rating 1: Immediate pleasure. Does this good or activity give you immediate pleasure? Is it fun, tasty, or pleasurable? Circle one of three responses: None immediate pleasure /Low immediate pleasure /High immediate pleasure Rating 2: Delayed benefits. Does this good or activity have benefits that last a few days? Is it healthy or educational? Circle one of three responses: None delayed benefits/Low delayed benefits/High delayed benefits Rating 1: Immediate pleasure? Espressino coffe: Espresso coffee: Chocolate Snack: Focaccia: Sandwich: Oransoda: Lemonsoda: Chinotto: Crostini: Salad: Fruit: Yoghurt low fat: Crisps: Cucciolone: Magnum Nuts: Cornetto: Caramelle: Textbook reading: Library research: Using gym equipment: Panorama Magazine: l’Espresso Magazine: Economy Magazine: Video games: Music on the web:

None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High

Rating 2: Long-term benefits? None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High None/Low/High

17

Debriefing Statement This was an experiment on consumer choice. We were interested in studying how people allocate their time among different activities. Thank you very much for participating. Please feel free to contact me at [email protected] at any time with questions, or if you would like to receive a copy of the research paper that results from this study. Thanks. ___________________________________ Enrica Carbone P.S. Feel free to encourage your friends to participate in this study. However, we ask that you not tell them any details about the actual tasks involved. We want them to come without pre-formed expectations. Thanks.

18

Appendix 2: the implications of alternative mappings from the indices to the weightings

Scenario 2: Consider the following mappings: Temptation coefficients Investment coefficients 1/2 0 1 2 0 1 2 0 0.000 0.000 0.000 0.000 0.125 0.250 1 0.125 0.250 0.500 0.000 0.250 0.500 2 0.250 0.500 1.000 0.000 0.500 1.000 With these we get the following weighted expenditures:

Temptation Investment Spot

58

42

Forward

59

41

Scenario 3: Consider the following mappings (note that expenditure is always allocated x% to temptation and (100-x)% to investment except when the good is rated zero on both indices): Temptation coefficients Investment coefficients 1/2 0 1 2 0 1 2 0 0.000 0.000 0.000 0.000 1.000 1.000 1 1.000 0.500 0.333 0.000 0.500 0.667 2 1.000 0.667 0.500 0.000 0.333 0.500 With these we get the following weighted expenditures: Temptation Investment Spot

62

38

Forward

65

35

19